This is the first study to explore the association between child and adolescent demographic, behavioral, sociocultural, attitudinal, and physical factors and physical activity behavior during the transition from adolescence to young adult. Young adults were categorized as persistently active, variably active or persistently inactive during this transitional life stage. Child and adolescent factors were more strongly associated with being persistently active into adulthood than variably active. The predictors varied according to gender and also physical activity category.
In our study, perceived sports competency was significantly associated with persisting with physical activity into adulthood for females. Males who assessed themselves as better than their same-age peers in sport were a little more likely to be variably active in adulthood, but not significantly more likely to be persistently active. Perceived competency refers to an individual's assessment of his/her own competency with respect to a task. Perceived competency is related to actual competency, but incorporates socioenvironmental factors such as feedback and reinforcement from significant others [27, 28]. For sports competency feedback also occurs via the outcome of participation (e.g. winning, finishing place, trophies). Assessment may be self-referenced (an individual evaluates their ability to perform the task independently of the performance of others) or other-referenced (an individual compares their ability to others) . In our study children's assessments of sports competency was other-referenced while recreational competency was self-referenced.
Previous longitudinal studies have found associations between teacher assessment of a student's ability (actual competency) in school sport or physical education in adolescence and physical activity in adulthood [13, 16]. Longitudinal studies in children and adolescents have shown an association between perceived sports competency and involvement in physical activity during adolescence [30, 31]. To our knowledge no previous longitudinal studies have examined the effect of perceived sports competency in childhood or adolescence on physical activity participation in adulthood. The current findings suggest that perceived sports competency is a particularly important factor for persistent physical activity by females. Improving perceived sports competency can occur through teaching specific skills to increase actual competency as well as by providing opportunities and supportive environments in which to develop or practise these skills .
In our study measures of physical fitness were associated with physical activity persistence by males and, less strongly, by females. For every minute taken to run the 1.6 km long run males were 15% less likely to be persistently active in adulthood and 6% more likely to be variably active into adulthood. In other words, the faster they ran the 1.6 km long run (indicating higher fitness) the more likely they were to be persistently active. The few studies that have examined the association between physical fitness in adolescence and physical activity in adulthood have found a positive relationship, although contrary to our study this relationship has been strongest for females and not males. These studies used measures of maximal oxygen uptake as well as endurance runs to assess cardiorespiratory fitness [3, 17]. Measures of strength were included in these studies, but they did not include the long jump and they did not incorporate measures of broad psychosocial, sociocultural or demographic factors.
Unexpectedly, not enjoying school sport was positively associated with being persistently active in males. There were only a small number of children in each physical activity category who did not enjoy school sport. The small number of boys who did not enjoy school sport but were categorized as persistently active (n = 8) were aged from 9 to 15 years. Five of them ran the long run faster than the mean for their age group suggesting they were fitter than average and rated themselves as better than their peers in sport (n = 4) and physical recreation such as swimming, tennis and skateboarding (n = 5). Enjoyment of sport can be associated with competition, social factors, skill development, competency, physical sensation and challenge or achievement . Given the small number of boys in this category and their age range it is difficult to explain this finding. It is possible that the type of activities offered for school sport were not activities at which they were proficient. Girls who did not enjoy school sport were less likely to be variably active into adulthood when compared with those who did enjoy school sport. The gender differences with respect to enjoyment of school sport may reflect differing sources of sport enjoyment.
Parental influences such as support and modelling have been identified as significant for physical activity in children and adolescents . In our study parental physical activity was not associated with physical activity persistence in females. Father's physical activity was significantly associated with persisting with physical activity during this transitional life stage in males. A review of parental influences on physical activity of children and adolescents found that the relationship between parental physical activity levels and child physical activity was equivocal . However, the majority of these studies were cross-sectional and not population based samples. The review did find evidence to support strong correlations between father and child physical activity levels, particularly fathers and sons. In a recent longitudinal study exploring the determinants of physical activity in adolescents aged 10 to 18 years in the USA mothers' physical activity was associated with physical activity in males and females . In this study fathers' physical activity was not measured.
Playing sport outside school was also found to be significant for males in our study. This positive relationship between extracurricular sports participation during childhood and adolescence and physical activity in adulthood has been found in other longitudinal studies [14, 16, 35, 36]. Where gender differences have been investigated this relationship has been found to be stronger for males than females . Involvement in sport outside the school environment during childhood and adolescence may help provide continuity during the transition from adolescence to adulthood when young people leave school.
The strong inverse relationship seen in our study between smoking and physical activity is consistent with findings from other studies that have investigated this relationship [8, 37]. Clustering of health promoting or risky behaviors is common and has been found in the same population of young adults . Smoking is commonly associated with more sedentary lifestyles although the mechanism for this is unclear . In this study the association for males and females suggested that smokers were less likely to be persistently active, but this was statistically significant only for females. The gender difference in the relationship between smoking and physical activity have been found in other studies investigating this relationship .
In our study, those females who had younger siblings were significantly less likely to be physical activity persisters into young adulthood. Our data do not provide insight into this finding. It is possible that it may reflect sociocultural expectations of female children to spend time caring for younger siblings thus providing less time for participating in physical activity, but this is conjecture. Studies investigating sibling relationships and physical activity have focused on assessing the relationship between physical activity participation among siblings . In one study exploring family structure and physical activity in children, data on siblings were categorized according to the presence of older siblings, total number of siblings and sex of siblings . The effect of younger siblings was not explored. Those children who had an older sibling were more likely to spend time in moderate intensity physical activity. This relationship was dependent on parental status.
Using the HLAQ, we found that one-fifth of young people persisted with leisure time physical activity between the ages of 15 and 29 years. The relatively small proportion of persistently active participants is consistent with other studies that have repeated measures of physical activity during this transition period [10, 11, 41]. Males were more likely to be persistently active than females; this difference has also been reported previously [4, 10, 42]. The largest physical activity category was the variably active. The variably active category included subgroups of young people who were increasingly or decreasingly active, but not persistently. However, statistical analysis of these separate subgroups (results not shown) provided no further insights than when the variably active category was considered as a whole. There were few childhood factors associated with being variably active, and physical activity participation for this group may be more strongly influenced by current life circumstances.
A limitation of this study is the use of the HLAQ that relies on participants' recall of their physical activity when they were aged 15-29 years. The mitigating factor is that we might expect accuracy of recall to be greater for organised leisure time physical activity than casual or incidental physical activity. Furthermore, organised physical activity is commonly of higher intensity than casual physical activity and in this study classification was based on time spent in moderate to vigorous physical activity and not on the other components that sum to total physical activity. Other studies have found that recall of physical activity over ten years is reliable [43, 44] and that the HLAQ is reproducible [20, 45]. The HLAQ has also been found to be predictive of bone mass in postmenopausal women . Measuring children's physical activity and attitudes by self-report is also challenging . However, the childhood physical activity self-report measures used in this study have been shown to be positively associated with the childhood cardiorespiratory fitness measures .
Another potential limitation of the study is the number of participants lost to follow-up. Compared to participants, non-participants were more likely to be male, smokers and classified as low SES; the effect of this loss to follow-up is not clear. The prevalence estimates we have reported for categories of physical activity persistence and other studies factors may not be generalizable beyond this sample, therefore. More importantly, however, the analytical associations reported should be robust because we have been able to explore associations with and adjust for a wide range of study factors, and none of those factors was so limited in distribution in this sample that estimation was infeasible.